Chapter V: Terms of Relationship (1)

Most descriptive monographs on primitive tribes contain lists of the words with which the natives designate their relatives by blood and marriage. The reason is far from obvious. Why should not this topic be left in the hands of a linguist-lexicographer? It is true that primitive usage in this regard is very quaint from our point of view, but so are primitive conceptions on a variety of subjects that likewise find expression in speech. The refinement of spatial distinctions in North American languages, the classification of colors or animals or other groups of natural phenomena are of equal intrinsic interest from a psychological point of view. Why, then, single out a particular department of the aboriginal vocabulary in a treatise on culture? The answer is simply this, that kinship terms have a direct relation to cultural data.

The very fact that primitive tribes frequently use terms of kinship as words of address where we should substitute personal names is a social practice of ethnological interest. But the essential point is that the terms used are often very definitely correlated with specific social usages. Generally speaking, the use of distinct words for two types of relatives is connected with a real difference in their social relations to the speaker. Thus, a majority of primitive tribes draw no distinction between the father’s sister’s daughter and the mother’s brother’s daughter. But among the Miwok of California, where one of the cousins may be married while the other is within the prohibited degrees, a discrimination is made in language. Again, in many regions of the globe an altogether special bond connects the maternal uncle with the sister’s son, and accordingly we find that he is very often sharply distinguished from the paternal uncle in nomenclature.

On the other hand, we can often explain very naturally the use of a single word for two or more relatives whom we designate by as many distinct words. The Vedda of Ceylon, for example, call the man’s father-in-law and maternal uncle by the same term. The reason is that here a man commonly marries his mother’s brother’s daughter; the mother’s brother _is_ his father-in-law, and this identity is expressed in the terminology. A different illustration is supplied by the Crow of Montana, who have one term for the man’s mother-in-law and his wife’s brother’s wife. The simple explanation is that both stand to him in the relationship of mutual avoidance, and it is this social fact that is expressed by the common designation. The same Indians apply the word for ‘father’ in a very inclusive manner, possibly to dozens of individuals; but closer examination shows that all of the people so addressed are entitled to the same kind of treatment by the speaker, to a peculiar form of reverence, and to a preferential rank in the distribution of gifts.

These few and casual examples possibly suffice to show why kinship terms deserve the ethnologist’s attention. Terms of relationship are, in some measure, indices of social usage. Where relatives whom other people distinguish are grouped together, there is some likelihood that the natives regard them as representing the same relationship because they actually enjoy the same privileges or exercise the same functions in tribal life. Where relatives whom other peoples group together are distinguished, there is some probability that the distinction goes hand in hand with a difference in social function.

Lewis H. Morgan, the pioneer in this domain of knowledge, was keenly alive to the social implications of kinship nomenclature. But while he endeavored to give an ultimate interpretation of it in terms of various social conditions, he was confronted with the fact that not every tribe had a terminology _sui generis_, but that nomenclatures of remote peoples were sometimes marvelously similar. Morgan boldly argued that such community of nomenclature established ultimate racial unity and on this ground coolly suggested a racial connection between the Hawaiians and the South African Zulu, between the natives of India and those of the Western Hemisphere.[1]

These speculations as to racial affinity have been rightly disregarded by later students, because to accept Morgan’s premises means running counter to the most obvious facts of physical anthropology. As Lubbock pointed out, we cannot assume that the Two-Mountain Iroquois are more closely akin to remote Oceanians than to their fellow Iroquois because some of their kinship terms resemble in connotation those of the Hawaiians. Nevertheless, Morgan was right in feeling that _some_ historical conclusions could be drawn from similarities of relationship nomenclature. We must simply bring this particular group of ethnological data under the same principle as other cultural phenomena. When the same feature occurs within a definite continuous region, we shall assume that it has developed in a single center and spread by borrowing to other parts of the area. When the same feature occurs in disconnected regions, we shall incline to the theory of independent development and shall inquire whether the course of evolution may have been due to the same cultural determinants, _i.e._, in this case, to the same social institutions.

After these preliminary remarks, we may turn to a closer scrutiny of the facts.

‘_Systems_’. Abstractly considered, it is conceivable that every individual relative might be designated by a different term of relationship by every other individual, just as each object in nature might theoretically be defined by some distinctive word instead of being placed in some such category as ‘tree’, ‘animal’, or ‘book’. Indeed, primitive people go rather far in their distinctions. Thus, in the Menomini family circle boys are not called ‘son’ or ‘brother’, but each is addressed by a word indicating the order of his birth, the oldest being ‘mudjikiwis’, the second ‘osememau’, the third ‘akotcosememau’, the fourth ‘nanaweo’.[2] But in this, as in every other department of language, economy has been exercised and instead of a chaotic number of distinct terms for every possible relationship, there is always a limited series, many distinct individual relationships being always grouped together under a single head. Thus, in English we apply the word ‘brother’ to a number of individuals regardless of their age relatively to ourselves or to one another and irrespective of the sex of the speaker. Yet, as the Menomini instance shows, the age distinction might very well have been expressed in speech and there are many Indian languages in which one set of terms is used by female and another by male speakers.

All the terms used by a people to designate their relatives by blood or marriage are jointly called their ‘kinship system’. This phrase is wholly misleading, if it is understood to imply that all the constituent elements form a well-articulated whole, for this probably never applies to more than a limited number of them, as will appear presently. But as a convenient word for the entire nomenclature of relationship found in a particular region the word ‘system’ may be provisionally retained. We may say, then, that systems of different peoples vary in their mode of classifying kin and it seems the ethnographer’s first duty to determine the types of system found and their geographical distribution.

At the present moment a satisfactory grouping of the world’s kinship systems is impossible, owing to our lack of knowledge of many areas. The task is also rendered very difficult by the frequent coexistence of distinct and even contradictory principles in the same ‘system’. Each of these may be defined separately, but to weld both or all of them into a unified whole defies our efforts. For example, the Masai of East Africa, in referring to the paternal uncle, simply combine the stems for ‘father’, _baba_, and ‘brother’, _alasche_, thus forming by juxtaposition of these primary terms the compound expression _ol alasche le baba_, which means literally ‘the brother of the father’. This mode of defining a relative’s status by combining primary terms of relationship or a primary term with a qualifying adjective as in our ‘grandfather’, is technically known as ‘descriptive’, and ethnologists are wont to speak of descriptive systems. As a matter of fact, this descriptive principle is highly characteristic of the Masai--but not when relatives are directly addressed by them. In such vocative usage, as it may be called, the father’s brother is called _baba_ like the father himself; the mother’s brother is not designated by a phrase composed of primary stems but by a new stem, _abula_, which is also used reciprocally for the nephew; while _koko_ serves to call both a paternal and a maternal aunt. These connotations introduce into the Masai ‘system’ a discordant principle by which relatives, instead of being defined descriptively, are grouped together in classes. But this ‘classificatory’ feature by no means characterizes all the vocative nomenclature. By far the majority of relatives are addressed by terms suggestive of the presents of live stock presented to them by the speaker; if the gift consisted of a bull, the word used is _b-ainoni_, from _oinoni_, bull; if an ass was given away, the vocative term is _ba-sigiria_, from _sigiria_, ass; and so forth. Accordingly, the vocative terms cited above are only employed by children, who have not yet presented stock to their kin.[3] In short, Masai terminology is molded by at least three entirely disparate principles.

We shall, accordingly, do well to amend our phraseology and to speak rather of kinship categories, features, or principles of classification than of types of kinship systems.

_The Descriptive Principle._ When we approach our subject in a purely empirical way, we are confronted with the fact that features do not, as a rule, occur sporadically but are distributed over continuous areas. Imperfect as is our knowledge of African systems, for example, we know that the descriptive feature of the Masai nomenclature does not appear everywhere, but flourishes especially among East African tribes, such as the Shilluk, Dinka, and other Upper Nile populations, and perhaps more widely where Arabic influence extends, the Arabian terminology being of a markedly descriptive character. In East Africa, indeed, there is almost quantitative proof of the dependence of kinship terminology on historical connection and geographical proximity. Among the Baganda, as among most Bantu Negroes, the descriptive feature is lacking and such a relative as the mother’s brother’s son, instead of being designated by a compound expression, is classed with the brother.[4] The Masai, who live surrounded by Bantu tribes, have a purely descriptive system for non-vocative usage but their vocative forms are in part classificatory, while some neighboring Bantu peoples have a correspondingly mixed system. The Shilluk and Dinka seem to use the descriptive principle exclusively, as do the Arabs. The Masai are undoubtedly closely allied with the Nilotes and markedly different from the Bantu. The conclusion is, therefore, inevitable that their terminology--whatever may be its ultimate _raison d’être_--is a function of their historical relations. They have descriptive features because they belong to a group of peoples of whom such features are characteristic. They have classificatory features because they have come into contact with peoples whose systems were characterized by such features and from whom they have borrowed them. The Shilluk lack the classificatory principle because they have not had the same alien influences as the Masai. The restriction of descriptive features to a definite part of Africa and their amalgamation with other features in the marginal section of this area show that kinship nomenclatures follow precisely the same rules as other elements of culture and that their distribution indicates probable or corroborates known tribal relations.

The descriptive principle is not restricted to East Africa and the Semitic family, but has been found in the Persian, Armenian, Celtic, Esthonian, and Scandinavian languages.[5] Although guesses might be offered, I do not feel that our present knowledge permits definite statements as to the historical relations suggested by the total range of the descriptive principle on the face of the globe.

_The Hawaiian Principle._ While the term ‘descriptive’ admits of a fairly unambiguous definition, the same cannot be said for the word ‘classificatory’. Morgan, after explaining his use of the former, states that a system of the second type reduces blood-relatives to great classes by a series of apparently arbitrary generalizations, applying the same terms to all the members of the same class. “It thus confounds relationships, which, under the descriptive system, are distinct, and enlarges the signification both of the primary and secondary terms beyond their seemingly appropriate sense.”[6] This is looking at the matter from the arbitrarily selected point of view of our own nomenclature (which Morgan improperly, as Rivers has shown, regarded as descriptive). Objectively considered, even descriptive terminologies are classificatory, inasmuch as they do not individualize, but content themselves with such generalizations as classing together, say, all the father’s brothers instead of uniformly specializing according to age. For this reason I regard as misplaced Dr. Rivers’ emphasis on whether a term designates a single individual or a wider group. What, then, lies at the basis of the classificatory principle? Dr. Rivers, following Tylor, reduces it to the clan factor or rather to the influence of the dual organization of ancient society, by which it was divided into exogamous moieties. But this important suggestion, to which we shall have to revert, applies avowedly only to one form of the classificatory system and involves, therefore, the hypothesis that this preceded other forms. This may prove to be valid, but we cannot prejudice an empirical survey by taking its proof for granted and cannot, therefore, simply substitute ‘clan’ for ‘classificatory’ systems--apart from the fact that to talk of systems instead of principles or features in this connection is demonstrably misleading.

It is quite clear that ‘classificatory’ can be used only in a loose sense, to indicate wider groupings of kin than those to which we are accustomed; and that there is no _necessary_ evolutionary relation between the two forms usually classed under this head. The empirical data are simply these. In certain systems, blood-relatives are classed according to generation regardless of nearness of kinship and of their maternal or paternal affiliations; in others, there is bifurcation, the maternal and paternal kin of at least the generations nearest to the speaker being distinguished. We may call the former the ‘unforked merging’, or geographically the ‘Hawaiian’ mode of classification; the latter may be correspondingly referred to as ‘forked merging’, or ‘Dakota’. One point which it is essential to remember even at this early stage of our survey is that these principles, together with the descriptive one, are very far from exhausting the varieties found.

Let us now consider the ‘unforked’ principle somewhat more closely as it finds expression among the Hawaiians. These people apply a single term, _makua_, to both parents and to all their parents’ brothers and sisters, sex being distinguished only by qualifying words meaning ‘man’ and ‘woman’. All related individuals of one’s generation are classed as brothers and sisters, certain distinctions being drawn according to the age of their parents relatively to that of one’s own parents and also according to the speaker’s sex, but none resulting from the differences in _nearness_ of kinship. The children of all these brothers and sisters are classed with one’s own children, and _their_ children with one’s grandchildren, while a single term embraces grandparents and all related members of their generation.[7] This age-stratification of blood-relatives with disregard of differences as to father’s or mother’s side occurs not only in Hawaii, but also in New Zealand, Kusaie, the Gilbert and Marshall Islands.[8] It is not uninteresting to note that Hawaii and New Zealand, though far removed from each other, coincide closely in other cultural features not shared with fellow-Polynesians, as Professor Dixon has recently shown in his treatment of Oceanian mythology. The geographical proximity of Micronesia to Hawaii hardly requires mention. Dr. Rivers points out[9] that certain Polynesian tribes in contact with Melanesians, whose systems display essentially the forked principle, _e.g._, the Tongans, use an intermediate nomenclature. We are thus again able to summarize the data in terms of historical connection. The assumption may be made that the ancient Polynesian terminology was that of Hawaii and New Zealand, which was modified where the Polynesians came into contact with diverse populations, and is shared by populations whose territory was presumably traversed by the Hawaiians. Dr. Rivers also states that the Burmese, Karen, Chinese and Japanese systems conform to the Hawaiian principle. He seems to depend on Morgan’s statement of the case, which may require revision. But, accepting the data as given and assuming that the Malay proper classify kin according to the unforked method, we should still have a perfectly continuous distribution for the Hawaiian features.

This would no longer hold if we accepted Morgan’s view that the Zulu of South Africa share the Hawaiian form, on which slender basis he advances the hypothesis that Kaffir and Polynesian have a common ancestry.[10] As a matter of fact, the Zulu nomenclature secured by Morgan does in some instances slur over the difference of paternal and maternal lines, to the exclusive dominance of the generation factor. Thus, man and woman call all the brother’s and sister’s children their sons and daughters without distinction, and the children of the father’s sister are classed with one’s brothers and sisters.

Nevertheless, even Morgan’s list reveals fundamental deviations from the Hawaiian principle. As he notes, the mother’s brother is _not_ classed with the father’s brother and father, and the assumption that he formerly was is mere guesswork. What particularly astonished Morgan, however, was that the father’s sister was not called mother, but father. This is, indeed, amazing, if we start from our own notions as to the necessity of distinguishing parental sex, and in addition assume that the Zulu system is a variant of the Hawaiian one. If we free our minds from these preconceptions, there is no mystery; the father’s sister is classed with the father simply in order to express the difference from the maternal line in accordance with the principle of bifurcation.

In order to gain greater clearness in this matter it is necessary to extend our investigation to other Bantu tribes, preferably to those whose territories approach that of the Zulu. The essential point to ascertain is whether paternal and maternal uncles and aunts are merged in one group or are distinguished.[11] Among the Thonga, who live north of the Zulu, the father’s sister, as in Zulu, is classed with the father, the word meaning literally ‘female father’ and thus emphasizing her separation from the mother’s side of the family. The Herero, according to Schinz, seem to class all aunts with the mother in vocative usage, but when not directly referring to these relatives they employ quite distinct expressions for the father’s and the mother’s sisters. In Baganda the difference between the two sides is marked. _Mange_ is mother, and the same word with the qualifier _muto_ means mother’s sister, while father’s sister is _sengawe_. Even clearer is the case for the maternal uncle. In the Ronga group of the Thonga he is called by a distinct word, _malume_, which almost coincides with Morgan’s Zulu term. In the Djonga division he is classed with the grandfather, not the father. By a quite distinct stem, the Herero sharply distinguish the mother’s brother from the father and his brothers. The same applies to the Baganda. As for the correlative term, from which Morgan infers that the Zulu once called the maternal uncle ‘father’, the Ronga have a distinct word for nephew, _mupsyana_, while the Djonga who class the mother’s brother with the grandfather consistently enough call the sister’s son ‘grandson’. Among the Herero, though uncles and aunts generally regard their nephews and nieces as their own children, the maternal uncle applies to them a distinct term, _ovasia_. Among the Baganda a man calls his son _mutabani_ or _mwana_, but his sister’s son is _mujwa_. I may add that the altogether peculiar bond of familiarity that links together mother’s brother and sister’s son[12] among some Bantu people is inconsistent with Morgan’s assumption that the relationships of maternal uncle and father were once grouped under a single head among tribes of this family, for as stated above, such specific social relationships are generally expressed by specific terms for the relatives.

The conditions obtaining within the speaker’s generation at first seem to lend some support to the conception of the Bantu system as dominated by the Hawaiian principle, since the terms for brother and sister are more widely employed by some Bantu than is compatible with the forked division of kin. But closer inspection proves that, whatever may be at the root of the Bantu classification, it is not the Hawaiian notion of marking off generations. Even in Morgan’s Zulu series, while a man calls his maternal uncle’s children by a special term, they address him as brother; that is to say, members of the same generation and sex are not all classed together. Among the Herero, where the children of a brother and sister (but not of _Geschwister_ of the same sex) regularly intermarry, they are placed in a category distinct from that of the children of two brothers and two sisters, who are one another’s brothers and sisters. In Thonga a boy calls his mother’s brother’s daughter ‘mother’, and she calls him ‘son’. To be sure, the Baganda draw no distinction between the brother, the father’s brother’s, the father’s sister’s, the mother’s brother’s and the mother’s sister’s son. On the other hand, only the father’s brother’s daughter and the mother’s sister’s daughter are a man’s sisters; his father’s sister’s and his mother’s brother’s daughter belong to the special category of _kizibwewe_, quite distinct from that of the sister, _mwanyina_.

To cut a long story short, all the evidence is opposed to Morgan’s assumption that the Bantu systems are patterned on the Hawaiian principle of grading relatives by generations. There are merely occasional suggestions of that principle which will be discussed below as to their theoretical bearing.

So far as I know, there is only one region of the globe outside of Oceania and the possible Asiatic range defined above, where a definitely Hawaiian classification of relatives by generations has been reported, _viz._, among the Yoruba of West Africa.[13] Unfortunately, no more recent check data for this section seem available. For another part of West Africa we have Mr. Northcote W. Thomas’ tables,[14] which reveal a rather perplexing condition of affairs that seems to demand intensive reinvestigation together with linguistic analysis. The principle of bifurcation seems to hold sway only in a very limited measure.

Thus, the Vai do not distinguish the father’s sister from the mother, though the mother’s brother is designated by a distinct term from that for father and father’s brother. Further, the term for child is extended also to brother’s child by both sexes contrary to customary ‘forked’ usage. But this cannot be interpreted as symptomatic of the Hawaiian principle since the sister’s child is designated by a special word, which, moreover, differs for men and women speaking. The Vai nomenclature is interesting in showing once more that a given ‘system’ is a complex growth that cannot be adequately defined as a whole by some such catchword as ‘classificatory’, ‘Hawaiian’, or what not. Not only do we find Hawaiian and Dakota elements in the same system, but even purely descriptive combinations of primary terms. Thus, the designation of the sister’s daughter’s husband is manifestly composed of the stems for sister’s child and husband, and a corresponding juxtaposition of stems results in the term for mother’s sister’s husband.

A similar phenomenon is presented by the terminology of the Timne, another Sierra Leone people. A superficial glance at the list suggests the Hawaiian principle: father’s brother and mother’s brother are grouped together, and so are the children of the maternal and the paternal aunt. But closer consideration shows that while uncles are classed together they are sharply separated from the father, that while aunts form a single group of _ntene_ the word for mother is _kara_ or _ya_, that there is no connection between the words for _Geschwister_ and cousins. In short, the Hawaiian generation principle does not apply.

What Mr. Thomas’ schedules from eight tribes illustrate once more is the overwhelming importance of historical, geographical and linguistic considerations. A cursory examination of the lists shows that not only the mode of classifying kin but the words themselves are identical in a number of cases in two or more tribes. Thus, _mama_ is grandmother in Karanko, Susu, Vai and Mendi. It is surely no accident that all of these belong to the same prefixless subdivision of the Sudanese languages: the similarity is due to historical relations. In some cases an identical word is shared by members of distinct subdivisions. Thus, the father’s sister is called _ntene_ not only in the non-prefixing Susu and Koranko speech, but also in the prefixing language of the Timne. A glimpse at Mr. Thomas’ map shows, however, that the habitat of the Timne adjoins that of both of the other tribes; a kinship nomenclature is, in a measure, a function of geographical position.

The last-mentioned term is suggestive in another way. Restricted among the Koranko and Susu to the father’s sister, it is applied by the Timne to the maternal aunt as well. Turning once more to the map, we discover that this latter mode of grouping, though not the same word phonetically, occurs among the Bulem, the immediate coastal neighbors of the Timne, who belong to the same linguistic subdivision, and also to the Mendi and Vai, to the east and southeast, who are members of the complementary subdivision. So far, this only indicates the spread of a terminological trait over a continuous area. But the data further suggest that the word _ntene_ may have been borrowed by the Timne rather than in the reverse direction, and that, as Mr. Thomas himself remarks, the Timne secondarily extended the term to include a maternal as well as a paternal aunt. This possibility is theoretically significant, first, because it indicates that Hawaiian analogies may develop independently of any such generation principle as dominates the Oceanian system; secondly, because it suggests that such simplicity of nomenclature, instead of being primitive as Morgan supposed, may represent a later development. To this point we shall have to revert later.

_The Dakota Principle._ Let us now turn to that principle which first aroused Morgan’s interest and which since his time has occupied perhaps more attention than any other, the classificatory principle _par excellence_ in Dr. Rivers’ opinion, which finds expression among such tribes as the Iroquois and Dakota. Like the Hawaiian principle, the Dakota alignment groups together, regardless of proximity of relationship, members of the same generation, but differs because in the speaker’s generation, the first ascending and the first descending generations, it separates the paternal and the maternal line. Another way of expressing the facts is to say that collateral and lineal kin are merged irrespective of nearness of relationship but with strict bifurcation of the parental lines. Thus, in Dakota[15] the father, father’s brother, father’s father’s brother’s son, father’s father’s father’s brother’s son’s son are all addressed _até_; the mother, mother’s sister, mother’s mother’s sister’s daughter are all called _iná_. So far we have a classing together of kin who in English are distinguished from one another. But there is separation of kin whom we class together, inasmuch as the mother’s brother is designated by a term distinct from that for father’s brother, _viz._, by _dēkcí_, and the father’s sister by a term differentiating her from the mother’s sister, _viz._, by _t ‘uwí_. Now, relationship is a reciprocal phenomenon, and accordingly we may expect that all those whom I class together under the term _até_ or _iná_ will address me by a correlative term. Actually, we find that the Dakota have a single word, _mi tcíñkci_, for son, brother’s son (man speaking), father’s brother’s son’s son (man speaking), etc., and for sister’s son (woman speaking), mother’s sister’s daughter’s son (woman speaking). To put the matter into our own speech, for the sake of simplification, those whom I call father and mother call me son. If logic shall prevail, the data hitherto cited involve the condition that the mother’s brother must not call his sister’s son ‘son’, but shall designate him by some distinct appellation correlative only with the term _dēkcí_; and this holds for the Dakota system where a man (not a woman) calls the sister’s son _mit ‘úncka_. Further this term is also used by a woman addressing her brother’s son, a point to which I shall have to return presently.

There are other logical implications in the features already mentioned. If the term for father embraces a number of other collateral relatives, we must expect a corresponding fusion of kin in the speaker’s generation. This is exactly what happens. Like many other primitive systems, that of the Dakota classifies brothers and sisters according to relative seniority and the speaker’s sex, but the same terms are applied to the other individuals who jointly designate the same members of the next higher generation as their fathers and mothers. In other words, a considerable number of cousins, irrespective of their varying degree, are classed with the brothers and sisters. But certain other cousins are _not_ so classed: they are the offspring of the father’s sister and the mother’s brother. Corresponding exactly to the fact that sister’s son (man speaking) and brother’s son (woman speaking) are denoted by a single word, we have the correlative phenomenon that the children of the paternal aunt and the maternal uncle are relatives of a special order, the boys calling one another _t ‘ahá ci_ and the girls _hà kā´ cí_, the girls calling one another _tcē´ pąci_ and the boys _citcé ci_.

In short, so far as the three middle generations are concerned, there is at least an approach to a real system--a unified logical scheme by which blood relatives are classified. If I am called father by a group of people, they are my sons or daughters; if I am their uncle, they are my nephews or nieces. In the former case, my sons and daughters are their brothers and sisters; in the latter my offspring are their cousins, with various refinements of nomenclature that are immaterial from a broader point of view.

The system is not perfect, because of the terminology applied to the offspring of cousins. As might be expected, a man regards the children of those cousins whom he classes with his brothers as brother’s sons, _i.e._, from the foregoing scheme, with his own sons. But contrary to what might be expected, he puts into the same category the sons of those male cousins designated by a distinctive term where we should expect a distinct correlative designation. Even Herr Cunow, who lays stress on the rational character of primitive relationship systems, is obliged to admit that there is inconsistency here.[16]

It cannot be too strongly urged that a given nomenclature is molded by disparate principles. It is, therefore, worth while to point out that the principle by which brothers and sisters are distinguished by seniority and the principle by which _Geschwister_ of the same sex use different designations from those of opposite sex have no functional relation whatsoever with the principle by which collateral and lineal kin are merged. Another trait of the Dakota system which is similarly independent of what I call the Dakota principle is the differentiation in stem for vocative and non-vocative usage or with the first, second and third person. Thus, the mother is addressed as _iná_, but ‘his mother’ is _hų´ ku_, from an entirely different root. Passing to the second ascending generation, we find a Hawaiian feature inasmuch as the principle of bifurcation no longer holds, grandfathers of both sides being designated by a common term. The Dakota case once more shows that, as Professor Kroeber long ago pointed out,[17] every system is in reality a congeries of systems or categories which must be analytically separated unless complete confusion is to result. There is no Hawaiian _system_, no Dakota _system_. But we can legitimately speak of the principle of generations and the bifurcation principle of merging collateral and lineal kin; and we can speak, by conventional definition of the geographical terms employed, of Hawaiian and Dakota features to express these and only these elements of the Hawaiian and Dakota nomenclatures.

To revert to the Dakota principle, as Morgan points out,[18] the same principle has in part molded the Iroquois system, and when we find that in addition to the logically related elements the apparently irrational classification of cousins’ offspring is likewise common to the two terminologies, the case for historical connection becomes very strong. This becomes a certainty when we find that in its essentials the principle finds expression in the system of the intermediate Ojibwa, while among other Algonkian tribes and among Siouan tribes other than the Dakota a marked variant from the Dakota type makes its appearance. In short, we have the Dakota principle spread over a continuous region, which is sharply separated from adjoining regions. It has, then, developed in a single center in this part of North America and has thence spread by borrowing.

If we ignore the mode of designating cross-cousins’, _i.e._, cousins who are children of a brother and a sister, and disregard certain other deviations constituting sub-types, we get a very much wider range of distribution for the Dakota principle in North America. The neglect of degree of kinship and the clear separation of the maternal and paternal line in the middle generations are features characteristic, probably, of the entire region east of the Mississippi and occur also in the Mackenzie River district, among the Tlingit and Haida of the Northwest Coast and most of the Plains tribes, in a part of the Pueblo territory (notably among the Hopi), and among the Miwok and adjacent populations in California. Since we are not by any means familiar with the kinship systems of the entire continent, it is necessary to supplement this statement with another indicating the regions where the Dakota principle is actually known to be lacking. The Dakota features are not found among the Eskimo, Nootka, Quileute, Chinook, various Salish tribes, the Kootenai, the Plateau Shoshoneans, nor in a large section of California to the north and east of the Miwok, and they are also absent from various Southwestern terminologies. The glib assumption of many writers that all of North America is characterized by a ‘classificatory system’ on the Dakota plan, is demonstrably false. The only reason for this belief is the historical accident that Morgan was conversant with the systems east of the Rocky Mountains and practically altogether ignorant of those of the Far West, and that since his time no one has systematically presented the data for what to him was a _terra incognita_.

Let us extend our search for evidences of the Dakota principle to other regions.

For Mexico, the data are not very satisfactory since we are obliged to rely on old Spanish sources and cannot be sure that our authorities were on the alert for differences from the familiar European nomenclature or always correctly represented what they did find. Thus, Dr. Paul Radin, who has kindly compiled for me a Tarascan list from Gilberti’s _Diccionario de la Lengua Tarasca_ (1559), finds the children of the father’s brother and of the mother’s brother classed with the son and daughter (contrary to the generation principle), but distinguished from the children of the father’s and mother’s sister. This would indicate a departure from both the Hawaiian and the Dakota scheme. A bare suggestion of the latter is found in a common term for father and paternal uncle. The Nahuatl data supplied by Molina in his _Vocabulario de la Lengua Mexicana_ (1571) show no difference between the paternal and maternal aunts and uncles. This does not apply to the Maya system reported by Beltran in his _Arte del Idioma Maya_(1742), but here the maternal and paternal uncle and aunt are not only distinguished from each other, but also from the father and mother, so that there is no merging of collateral and lineal lines in this generation. Accordingly, it is somewhat surprising to find that the children of a brother are classed with one’s own children (male speaking?) and that a woman applies the same term to her sister’s children, in accordance with Dakota usage. A very interesting feature of the Maya nomenclature is that differences in generation are conspicuously ignored in several instances. The paternal grandfather is classed with the elder brother, a single reciprocal term is used for daughter’s son and mother’s father, one word denotes the son’s son and the younger brother.

For Central and South America the data, from a cursory inspection, seem somewhat more adequate, though we must eagerly await a more thorough-going survey of this region than can at present be offered. The Miskito of Nicaragua call the mother’s sister _yaptislip_, which is merely a modification of _yapti_, mother, but while the father’s brother, _urappia_, is classed with the step-father, he is distinguished from the father, _aisa_. At all events, there is a distinctive term for maternal uncle, _tarti_, and correlatively a special designation, _tubani_, for the sister’s son (man speaking). For the father’s sister our authority gives only a descriptive term: _saura_ may be the correlative term, but it is simply translated ‘brother’s child’. Of the four terms for cousin, one is descriptive (child of brother or sister), two coincide with the regular words for _Geschwister_, the fourth is unfortunately not clearly defined so that its application to the cross-cousin, which would conform to Dakota usage, remains problematical. The terms of affinity are interesting inasmuch as the principle of reciprocity appears here. Thus, _dapna_ means both father-in-law and son-in-law, and the same descriptive expression, oddly enough, is applied to the mother-in-law and daughter-in-law in female speech.[19] The former instance of reciprocity recurs among the Chibcha of Colombia and we may thus have here another case of the geographical localization of kinship features. The Chibcha list supplied by one of Morgan’s informants,[20] imperfect though it is, records some suggestive facts. The term for father’s brother seems only a variant of the word for father, and is clearly distinct from that for maternal uncle. The designations for both kinds of aunt are doubtful. In the speaker’s generation ‘parallel’ male cousins, _i.e._, the sons of two brothers and of two sisters, are grouped with brothers and distinguished from cross-cousins, as they are in the Dakota system. That a woman calls her father’s sister’s son by the same term as her husband is a fact of some theoretical importance since it suggests the possible occurrence of cross-cousin marriages.

From Martius’ rather confusing Carib list we may reasonably infer that the paternal uncle was classed with the father in male speech and distinguished from the mother’s brother. One of three terms used by a man in designating his son coincides with that applied to a brother’s son, but differs from the word applied to the sister’s son. These are Dakota features; and the peculiar statement that children of sisters were allowed to marry while those of brothers were not, coupled with the remark that _Geschwisterkinder_ call one another brothers makes us suspect that we have here merely an abortive attempt to describe the difference between parallel and cross-cousins recognized on the Dakota principle. The Tupi terminology furnished by the same writer does not suggest the bifurcate feature. Though a single word denotes the father, his brother and other paternal kinsmen, it seems to extend likewise to the corresponding relatives on the mother’s side. In the second ascending generation the grandfather’s brothers and male cousins are classed with the grandfather--a Hawaiian trait if both sides of the family are meant to be included, but one common to most systems on the Dakota plan for the middle generations.[21] From the third great South American family I can get no satisfactory evidence of bifurcation on the Dakota plan. According to an accessible glossary of various Arawak tongues, the Siusi is the only language that discriminates between the paternal and maternal uncle, and even here the former is also distinguished from the father, so that there is no merging of collateral and lineal kin. Similarly, the word for aunt is different from that for mother; and here the principle of bifurcation is completely discarded, since a single word denotes father’s and mother’s sister.[22]

Bifurcation may be a dominant feature of systems which nevertheless differ markedly from the Dakota nomenclature because of their demarcation of collateral and lineal kin. Thus, the Araucanians of Chile call the father _chao_, the father’s brother _malle_, the mother’s brother _huecu_; the mother is _ñuque_, her sister _ñuquentu_, the father’s sister _palu_.[23] Here the designation of the maternal aunt is clearly derived from that of the mother but we cannot tell whether this merging is an ancient feature which appears in other parts of the system or a recent development. We learn from another source that the brother’s sons are differentiated from the sister’s,[24] but unfortunately there is no statement as to whether the former in male speech and the latter in female speech are classed with one’s own sons.

Bifurcation without reduction of the collateral lines is characteristic of the system of the Sipibo, who inhabit the country about the Ucayali River. Here the father is _papa_; the father’s brother _eppa_, the maternal uncle _cuca_; the mother _tita_, her sister _huasta_, the paternal aunt _yaya_, and of the three words for brother’s son (_pia_, _nusa_, _picha_) none even remotely resembles that for son, _baque_.[25]

To sum up the facts hitherto cited. If the doctrine of the unity of the American race depended on the uniformity of kinship terminologies in the New World, it would have to be mercilessly abandoned. Meager as are our data for the area south of the United States, we can find positive indications of nomenclatures with Dakota features only among the Caribs and the Chibcha, with occasional suggestions elsewhere. The Tupi and Arawak systems are markedly unforked; the Araucanian and Sipibo terminologies are forked but non-merging. Taking into account the large section of North America already defined as lacking bifurcation with merging, we thus have an immense territory in America in which the Dakota principle does not occur.

But, as the African facts cited above show, the Dakota principle is not confined to a portion of the Western Hemisphere. It is impossible completely to define its distribution in various parts of the globe, but the main regions must be indicated. As Morgan pointed out on the basis of Rev. Fison’s information,[26] the principle occurs in the nomenclature of the Coastal Fijians, and corroborative evidence has recently been furnished.[27] Rivers has shown that the typical Dakota principle appears in other parts of Melanesia, often with a very interesting additional feature in the designation of cross-cousins, who are not only rigidly distinguished from the parallel cousins but classed simultaneously as brothers-in-law and sisters-in-law, _e. g._, in Guadalcanar.[28] Bifurcation with merging of collateral and lineal relatives also characterizes at least some of the terminologies of New Guinea.[29] The same certainly holds for a large portion of Australia, though almost everywhere certain local refinements are apparent. Thus, the Urabunna apply one term to the father and the father’s brothers, as might be expected. But instead of merely separating the mother’s sisters from those of the father by grouping them with the mother, there is an additional dichotomy into the mother’s elder sisters, _luka_, who are classed with the mother, and the mother’s younger sisters who are differentiated as _namuma_. Corresponding differentiation occurs in the speaker’s generation, where the father’s elder sister’s daughters are distinguished not only from parallel cousins but from the father’s younger sister’s daughters. Nevertheless, the essentials of the Dakota principle are manifest.[30]

Here it is worth while to point out again how misleading it is to treat accidentally associated features of a given system as functionally correlated. The Urabunna system, like that of other tribes, is not an organically unified whole. Thus, over and above the usual trait of bifurcate merging, we find the feature that a grandparent and grandchild use a common term in addressing each other. This reciprocity is often referred to as characteristic of ‘classificatory systems’. It is nothing of the kind. In North America it occurs precisely in systems lacking the classificatory principle altogether. Apart from this, there is no manifest connection between the principles of grouping together relatives of alternate generations and the principle of classing under one head relatives of the same generation and side of the family. The mere fact that kinsfolk are united whom we happen to separate in nomenclature is a purely negative and insufficient reason for postulating an essential relationship between two modes of classification.

Finally, there are a number of Asiatic tribes whose systems reveal the essentials of the Dakota principle. At least a close approximation occurs in the nomenclature of the Gilyak of the Amur River country, where, except for the grouping together of father’s and mother’s sister, the two parental lines are kept apart while on either side the customary merging takes place.[31] The system of the Tamil, as Morgan emphatically pointed out, is almost identical with that of the Seneca Iroquois.[32] The essential resemblance to this type of the Toda,[33] Singhalese and Vedda[34] terminologies has since been established.

We are here again confronted by a problem in distribution that does not differ in principle from ethnological problems relating to other phases of culture. A sharply individualized feature is found not like the Hawaiian principle practically within the limits of a single continuous area but in several diverse and remote regions of the globe. It is impossible to hold with Morgan that the similarity found is an index of racial affinity unless we are willing to assume that the Indians of the eastern United States are not related at all to those west of the Rocky Mountains. The principle of diffusion obviously accounts for much. No one would hesitate to assume that the Singhalese and Vedda systems are connected and we should willingly regard both as historically related to the nomenclature of southern India. We might even be willing to grant that the Melanesian and Australian variants of the Dakota principle had the same source of origin. But how can we explain the predominance of the identical principle precisely in the eastern regions of North America and its absence in a great part of the Far West? And how can we account for the African approximations to the same pattern? We seem to have an independent evolution of the same highly characteristic trait in at least three distinct areas. Must we content ourselves with simply accepting the data as irreducible ethnological phenomena or can we carry our analysis a step further?

That the inclusiveness of terms which strikes us in the systems sharing the Dakota principle is somehow connected with the social divisions of the tribes concerned has been repeatedly noted. Even in his earlier, purely descriptive work Morgan remarked that among the Iroquois clan members were brothers and sisters as if children of the same mother.[35] Similarly among the Tlingit we are told that a single word is applied to the mother’s sister and all other women of the same moiety and generation.[36] The Yakut apply one term to any woman older than the speaker and belonging to the same gens.[37] Such instances might easily be multiplied. It is therefore rather natural to look to a clan or gentile system for the explanation of the ‘classificatory feature’, _i.e._, of bifurcate merging.

This hypothesis, which has recently been discussed by Swanton,[38] was already advanced only to be proved inadequate by Morgan himself. Taking the Seneca for illustration, where descent is in the maternal line, Morgan shows that the children of two sisters would indeed be members of the same clan and hence clan brothers and sisters but that this explanation no longer holds for the children of two brothers. By the law of exogamy these would be required to marry into another clan and there is no reason why their wives should belong to the same clan. Hence the brothers’ children will not be clan brothers and sisters, yet, according to Seneca terminology, the offspring of brothers no less than of sisters are classed with own brothers and sisters. Accordingly, the clan system--though it has a definite place in Morgan’s scheme of evolution--is not regarded by him as the determining factor of the Seneca-Dakota principle.[39]

But the objection vanishes if we accept the theory that the Dakota principle arose as a reflection not of a multiple clan system but of an organization with exogamous moieties. This theory, which to my knowledge was first developed by Tylor[40] and has since been advocated by Rivers,[41] has obvious advantages. Even on the simple clan hypothesis it is clear why the father’s brothers should be classed with the father and separated from the maternal uncles, since the latter by exogamy must belong to a different clan. The term which we translate ‘father’ would really be seen to mean ‘male member of the father’s clan and generation’. With the moiety theory the same facts are explained, but also in addition the designations for other relatives. To take again the Seneca instance, the sons of two brothers _must_ be members of the same social division because with a dual organization the brothers are restricted to the same division in the choice of a mate; hence it is quite natural that the sons of brothers should call one another brothers. Again, the difference between parallel cousins and cross-cousins is perfectly intelligible. The mother’s brother’s and the father’s sister’s son can never be of my moiety; if descent is matrilineal they belong to my father’s moiety, if patrilineal to my mother’s. Hence it is natural that they should not be classed with my brothers who in either case are my moiety-mates. This hypothesis also explains features not yet referred to, but often found in conjunction with those grouped under the heading of the Dakota principle, _e. g._, the frequent classification of the father’s sister’s husband with the maternal uncle. Given exogamous moieties, these relatives must belong to the same half of society, to my own moiety if descent is maternal, to my mother’s if it is patrilineal. The Tylor-Rivers theory thus explains very satisfactorily the rather numerous features that jointly constitute what I have called the Dakota principle; we can at once see that here is not an arbitrary rule of classification but a definite rationale.

However, it is worth noting that while the moiety theory explains a number of traits better and more simply than the hypothesis of multiple clans or gentes of which it is a special form, the latter is not in so bad a plight as Morgan would have us believe. That I should call my father’s brothers and male cousins of the paternal line ‘father’ and my mother’s sisters and female cousins of the female line ‘mother’, follows from the general hypothesis of exogamy no less than from the moiety theory. The difficulty urged is the grouping together of brothers’ sons who are not clansmen under a matrilineal organization with sisters’ sons who are. But all terms of relationship are correlative: the concept of elder brother is meaningless without the correlated concept of younger brother; so the very fact that I address my father’s brother as ‘father’ has as a necessary consequence that he should address me as ‘son’ regardless of whether his own son is in my clan. Similarly, the fact that my father’s brother’s son and I both address my own father as father makes us brothers irrespective of clan affiliation. Clan affiliation is still the primary determinant since it fixes the connotation of the word translated ‘father’, while the other usages mentioned are derivative applications. The objection that naturally obtrudes itself is why the term for father should be taken as the starting-point rather than that for son or brother. The answer lies in the fact that in a number of instances the term for father has an emphatically clan or gentile significance, being extended even to father’s clansmen of the speaker’s generation, as among the Crow and Arizona Tewa. Nevertheless, it cannot be denied that from the point of view of summarizing the data comprised under the caption of ‘Dakota principle’ or intimately linked with them the moiety theory is distinctly superior. Thus, the union of father’s sister’s husband and mother’s brother under a single head does not follow from a multiple clan or gentile organization but is intelligible on the basis of a dual division.

The weakness of the moiety theory lies in another direction. In order that the dual organization may fashion kinship nomenclature, it must of course exist. Now it does occur in Australia and Melanesia, though not universally, and in part of North America, but it is lacking in many regions of this continent and, so far as I know, in Africa. If we derive the Dakota principle exclusively from the dual organization we are therefore obliged to assume either that this institution once had a far wider range of distribution or that the nomenclature it produced traveled independently of the moieties to a considerable number of other peoples. This is a difficulty that must be frankly recognized.